A few "reality checks" with ADCs :

• The precision of the ADC is limited to 0.1% at full scale (1024). This means that when measuring smaller values, it's less precise. For example, when measuring 1/10 of the reference, your precision is down to 1% already.
• Analog noise is usually in the range of 10mV is your circuit is carefully designed. Similar to above, it has a bigger effect on small values, but also on smaller references (1.1V is 3x more sensible to noise than 3.3V). Add a 10nF capacity in parallel to your 330k for better resilience.
• ADC is slow and needs time to calculate. Wait enough time between changing parameters and reading value (the Ardunio routines don't).

I'd suggest you do a for() loop of 20x reading your ADC, and see the variation : mean(), std() and trend (for example last 5 - first 5). It could learn you a few things.

@rozpruwacz I can guarantee you that the IDE will NOT link all the libraries in every project. I have many, and they would never fit in one 328P memory. Even the parts of libraries that are not used do not get in.

That said, there are many "memory eaters" in libraries, so you should only consider using one if you really need it.

A second point of attention is using floats. They are deadly for memory, because the 328p is not able to calculate with floats natively. Division is already huge, but if you use any sin or tan, kB are used.

A third eater is ... objects. When you create an object (from a library), every part of it is created. So if it done according to the good old rules, it starts bay uselessly copying the parameters of the constructor into its internal variables ... that maybe won't ever change.

Strings are also eating a lot of memory. Both in creation and in manipulation.

To see what eats your memory, start from an empty sketch setup+loop, and add each part of your sketch (in a compilable way) step by step. You could be surprised of where the biggest memory eaters are.

@zboblamont I also have a quite complex and too expensive system (heatpump, solar panels, RF devices) for my (floor) heating, and indeed you did the right thing : insulation is the key. Your (and every user's) goal is to spend less, Honeywell's goal is to sell you their stuff (sales) and lead you into believing that it was the right choice (marketing). So if you want to get additional benefits, you can also ... double check your air tighness and ventilation, and monitor your electricity bill. Not sure if PV is an interesting alternative where you live.

One of the reason of the lack of feedback is that your OP is not really a (simple) question. Not sure what you ask opinion on, and you already did your homework. A second reason is ... summer.

A few "reality checks" with ADCs :

• The precision of the ADC is limited to 0.1% at full scale (1024). This means that when measuring smaller values, it's less precise. For example, when measuring 1/10 of the reference, your precision is down to 1% already.
• Analog noise is usually in the range of 10mV is your circuit is carefully designed. Similar to above, it has a bigger effect on small values, but also on smaller references (1.1V is 3x more sensible to noise than 3.3V). Add a 10nF capacity in parallel to your 330k for better resilience.
• ADC is slow and needs time to calculate. Wait enough time between changing parameters and reading value (the Ardunio routines don't).

I'd suggest you do a for() loop of 20x reading your ADC, and see the variation : mean(), std() and trend (for example last 5 - first 5). It could learn you a few things.

@zboblamont I also have a quite complex and too expensive system (heatpump, solar panels, RF devices) for my (floor) heating, and indeed you did the right thing : insulation is the key. Your (and every user's) goal is to spend less, Honeywell's goal is to sell you their stuff (sales) and lead you into believing that it was the right choice (marketing). So if you want to get additional benefits, you can also ... double check your air tighness and ventilation, and monitor your electricity bill. Not sure if PV is an interesting alternative where you live.

One of the reason of the lack of feedback is that your OP is not really a (simple) question. Not sure what you ask opinion on, and you already did your homework. A second reason is ... summer.

@alowhum Zigbee being "a standard" is not an advantage at all. As you put it, it is indeed not true at all. Using "a standard" in the meaning of "a formal definition of one of the piece of your puzzle" is only an advantage when you want to potentially replace that piece by a piece produced by another manufacturer and still get that talking to the rest of the puzzle.

You should also consider that there are two kind of standards : "committee" standards, where experts provide a formal document and people pay to adopt that way of doing things, and "de facto" standard, where there are enough users that do things the same way (or buy the same thing) that it guarantees sufficient resiliency. For example, USB is a committee standard (but if you use A, micro, mini etc. you have to adapt), Arduino Uno is a de facto standard (and there as well, you can choose between Uno or micro, etc.), and its success made possible the clones costing a few bucks.

@mhkid For battery powered sensors, you should always try to go 3V3. And if necessary change the model of sensor. At 3V3 the consumption is lower as well, so you win twice the energy (very roughly). Your device should also sleep continuously between the sensor captures. It can become tricky. But I used a DHT22+328p+TI CC RF sensor powered by 3V3 on a single AA cell and up-converter for 5 years now, and I only need to recharge it every 4-6 month.

@rozpruwacz I can guarantee you that the IDE will NOT link all the libraries in every project. I have many, and they would never fit in one 328P memory. Even the parts of libraries that are not used do not get in.

That said, there are many "memory eaters" in libraries, so you should only consider using one if you really need it.

A second point of attention is using floats. They are deadly for memory, because the 328p is not able to calculate with floats natively. Division is already huge, but if you use any sin or tan, kB are used.

A third eater is ... objects. When you create an object (from a library), every part of it is created. So if it done according to the good old rules, it starts bay uselessly copying the parameters of the constructor into its internal variables ... that maybe won't ever change.

Strings are also eating a lot of memory. Both in creation and in manipulation.

To see what eats your memory, start from an empty sketch setup+loop, and add each part of your sketch (in a compilable way) step by step. You could be surprised of where the biggest memory eaters are.

@janerik I'm not sure what you mean by "a CC110x module without RF stage", as the CC110x is the RF part (for panstamps as well as other systems). Do you only use the driver or do you use panstamp HW as well ? Do you use SWAP ? I've been using panstamps with SWAP for more than 5 years now, indor but in a humid environment (pool), and I didn't notice this. My DHT22 sensor (on "sensor" board, going 6 month with a single AA cell !) stopped measuring humidity a few months ago, but the board (and the temp part) is still working without any hiccup.

@alexsh1 You didn't say which direction is too big. But another smaller alternative is also panstamp NRG. a bit pricey but it comes with its own RF component. I was actually thinking of asking about its support in another thread...

@alexsh1 Wouldn"t a teensy 3.sth fill all the requirements ? Or an ESP8266 if you don't need ADC mux. ESP32 is you do.

@kayedee said in How do I access Mysensors Cloud:

@hek
OK, but you have not replied to the remainder of my post, am I to assume that the cloud is not going to happen?

@hek, thanks for the incredible work done at mysensors. Do you have any update after three years on that topic ? Is there going to be a MySensor cloud someday ?